December Books Solid, February Sits Empty — the Benefit-Reset Whiplash 2026 Demand Models Can Finally Forecast

Run one report tonight before you lock up: active patients with remaining insurance benefit as of December 1. In a practice with 1,900 active patients you will usually find somewhere north of 700 names, and behind those names, more than half a million dollars of annual maximum that expires at midnight on the 31st. Almost none of it gets spent. Then, six weeks later, your hygiene columns have holes you could drive a van through and your scheduling coordinator is calling the ASAP list at 8:15 in the morning trying to fill a 10 a.m.

Dentistry does not have a demand problem. It has a demand shape problem, and the shape is the same every year: a benefit cliff on December 31, a deductible reset on January 1, a dead February, a sealant-and-mouthguard bump in August, and ortho starts clustered in June when school lets out. Everyone in the practice knows the pattern exists. Almost nobody staffs, orders or prices against it, because the pattern has never been written down as a number.

The whiplash, measured in operatory hours

Take a six-operatory general practice with three hygiene columns. In the first three weeks of December, hygiene runs over — double-booked, lunches eaten standing, the doctor doing exams between crown preps, and the restorative schedule full of people who suddenly want that crown "before the end of the year." Utilisation is effectively above capacity, so the practice turns work away or pushes it into January, where the patient's deductible has reset and the case dies.

Then February. Two of the three hygiene columns are running at maybe seven patients instead of nine. Your hygienists are salaried or guaranteed hours, so you pay for the empty chair either way. The office manager fills the time with recall calls, which is the polite way of saying the practice is now paying skilled clinicians $45 to $60 an hour to leave voicemails.

What the office does today: last year's schedule, plus a feeling

The honest description of dental capacity planning is this — the office manager looks at what happened last year in the same month, adjusts for whichever hygienist is pregnant or leaving, and books the temp agency when it feels tight. Lab and supply ordering works the same way: an implant case comes in, someone calls Patterson or Schein, and the December crown rush is absorbed by paying rush lab fees.

The reason it stays a feeling is that the underlying history is genuinely awkward. Each patient is on a six-month or three-month recall clock. Each has a different plan with different frequency limits and a different renewal month — plenty of plans renew on the employer's fiscal year, not in January. Cancellations cluster around weather and school holidays. Classical forecasting methods want long, clean, evenly spaced history, and dental history is sparse, spiky and full of exceptions. So the spreadsheet gets abandoned and the feeling wins.

flowchart TD
  A["Recall due dates from Dentrix"] --> D["Weekly demand forecast by column"]
  B["Remaining benefit and plan renewal month"] --> D
  C["Three years of cancellations, weather and school breaks"] --> D
  D --> E{"Forecast vs. scheduled hours, six weeks out"}
  E -->|Short by more than 12 hours| F["Open a Saturday, book the temp hygienist now"]
  E -->|Long by more than 12 hours| G["Pull the February recall wave forward, offer it in December"]
  F --> H["Office manager approves the staffing change in the huddle"]
  G --> H

What changed in 2026 about forecasting lumpy demand

Demand forecasting has quietly become one of the most-adopted business uses of AI anywhere — roughly 48% of manufacturers now run it, and pricing optimisation sits around 72% in retail and e-commerce. Those two numbers are the interesting part, because manufacturing and retail are not more sophisticated than dentistry; they simply got tools that tolerate messy history first.

A demand forecast, in a dental practice, is a per-week estimate of how many hygiene and restorative hours patients will actually want — built from your own recall clock, plan renewal dates and cancellation history — accurate enough to staff and order against six weeks ahead.

What is different in 2026 is that the models no longer need you to clean the data first. You can hand over three years of exported appointment history with its gaps, its COVID-era holes, its duplicate patient records and its inconsistent provider codes, and get a weekly forecast per column back. And the cost of running it collapsed — frontier models are roughly ten times cheaper than they were in 2025 — so a practice can re-forecast every Monday morning instead of once a year, which is the difference between a planning document and an operating tool.

The other half nobody does: what your fees actually are

Forecasting tells you when the chairs fill. Pricing tells you what an hour in them is worth, and this is where most owners have simply stopped looking. Two questions to sit with:

• When did you last move your office fee? Not the PPO schedule — your own full fee, the one that sets the write-off on every in-network claim and the price for every cash patient. Many practices raise it "about 3%" across the board every couple of years, which means individual codes drift far from where the market sits in your ZIP.
• What is your in-house membership plan priced at, and does it clear its own cost? A plan at $348 a year covering two cleanings, two exams and bitewings is either a smart way to fill February or a way to sell $520 of hygiene for $348, depending on how many members actually come twice.

The same models that forecast demand will price by code against your own accepted-fee history, your PPO schedules and your local market data — and, more usefully, will tell you which twelve codes actually matter. In most general practices, a handful of codes carry the arithmetic: the adult prophy, the periodic exam, bitewings, the crown, the buildup, quadrant scaling and root planing, and periodontal maintenance. Everything else is noise you can leave alone.

The arithmetic: one February, one temp hygienist, one December Saturday

Assumptions, all illustrative and all replaceable with your own: three hygiene columns, 8 patients a day each at capacity, 200 clinical days a year, average hygiene visit producing $210 including the doctor's exam and radiographs, hygienist cost $52 an hour fully loaded, temp agency $71 an hour.

Add a fee review that moves eight codes to where your market actually sits, and on 1,900 active patients the fee side is usually larger than the scheduling side. But the scheduling side is the one you can prove within a single quarter, so start there: keep the forecast, keep the actuals, and compare them in April.

Where the forecast will be wrong, and what you should not automate

Three failure modes worth knowing before you trust a number.

First, the forecast cannot see a new employer moving into town or the industrial plant that just cut its dental benefit — the two events that actually change your year. Those arrive as gossip at the chamber of commerce meeting, not as a data point, and the model will keep forecasting last year's employer mix until you tell it otherwise.

Second, do not let anything set prices automatically. Fee changes touch your PPO participation, your written financial policy, and in some states your posted-fee obligations. A model can propose a fee schedule; the owner and the office manager approve it, and it goes in once, on a date, with the front desk briefed. Silent per-patient pricing in healthcare is a bad idea for reasons that go well beyond the arithmetic.

Third, a forecast that says "you will be short 14 hygiene hours" is worthless if you cannot hire. In a lot of markets the constraint is not demand, it is that there is no hygienist to hire at any wage. The forecast then becomes an argument for a different lever — extending hours for the hygienist you have, or moving recall intervals, not filling a column that will never be filled.

Frequently asked questions

Can this read straight out of Dentrix or Open Dental?

You do not need a live connection to start. Export appointment history, recall due dates and production by code to a spreadsheet — most practice management systems will do three years at once — and work from that. Practices on Open Dental have the easiest time because the database is directly readable; Dentrix, Eaglesoft and Denticon users typically start with scheduled exports. Get value from the export before you pay anyone for an integration.

Is this useful for a single-location practice or only for a DSO?

A group with 22 locations gets the bigger dollar answer, because a regional operations manager can move a hygienist between two offices eight miles apart. But a single office gets a cleaner answer, because one location's history is not muddled by different fee schedules and different payer mixes. The lever is different — a solo owner adjusts hours and recall waves rather than moving staff.

Won't pulling recall forward into December just empty out the spring?

Only if you pull the wrong patients. The candidates are people whose recall falls in the first two weeks of January anyway and who still have benefit left this year — moving them three weeks earlier gets them a covered visit and clears space in your slowest month. Patients due in April stay in April.

How long before I know whether the forecast is any good?

One quarter, if you write the forecast down before the quarter starts. Keep a single sheet: forecast hours per column per week, actual hours, and the variance. If the variance is inside about 10% by week six, staff against it. If it is not, you have found a data problem worth fixing, which is also useful.

One practical note on the recall wave itself: pulling 40 patients forward into December means 40 outbound conversations in a month when the front desk has no spare minutes. CallSphere builds AI voice and chat agents that answer the practice line and the web chat around the clock, book and reschedule appointments and capture new-patient enquiries — which is usually how a forecast turns into a filled column rather than a report nobody had time to act on.